Physics-governed bounds
Model predictions stay within limits dictated by hull, propeller, and engine physics — not just data patterns.
Technology — 01
Physics-Informed AI
Hybrid models that embed hydrodynamic and thermodynamic physics into machine learning, so predictions stay bounded by real marine engineering laws.
Rather than relying purely on historical black-box patterns, our models are built on Holtrop-Mennen resistance theory and MAN B&W engine physics, combined with gradient-boosted learning models and live marine weather data. This grounding lets the system extrapolate safely to conditions it has not seen before, instead of guessing from past patterns alone.
Architecture
Sparse-data robustness
Physics anchoring keeps predictions stable even where operational data is limited.
Root-cause diagnostics
Deviation from physical baselines is used to isolate whether loss stems from hull, propeller, or engine deterioration.
Development evidence
294+training voyage rows — MT Magenta Ray
107operational features per data row
100%route position data coverage
6-cyltwo-stroke ME engine model
Tools & methods
Discuss this technology
Talk to our engineering team about modelling, simulation, or intelligent systems work.